MetaboAnalyst5.0 Functional analysis p-value issue

MetaboAnalyst
1

I encountered some issue when using ‘MetaboAnalyst5.0’ to do functional analysis(Input data type ‘MS peaks’), it seems like the -log10(P-value) showed on the graph does not match the mummichog p and GSEA P in the table below, this issue exists for the datasets of my own and also with the example datasets from the website. I attached the results of using one of the example data(IBD). For example, the graph shows -log10(p) for mummichog is 2.6, but from the table, mummichog p-value is 0.1022, -log10(0.1022)=0.9905.

This p-value issue only happens if I choose ‘Mummichog’ and ‘GSEA’ at the same time in the “set parameter” step, if I choose only one of them to run the analysis, the p-value matches the -log10 in the graph. Are the p-value transformed in some way before log transformation when using both mummichog and GSEA?

image
image902×491 53.2 KB

image
image869×448 43.6 KB

2

Thanks for your report and careful checking. The -log(p) values in x-axis and y-axis are scaled values between 0-4. That’s why they are different from the raw p value. The floating label in scatter plot will be updated soon to indicate this.

3